Second case of $@ is not checked to be a cell.

[eamsden]

Describe the bug
$@(%our ship) compiles

To Reproduce
Compile |% +$ wat $@(%our ship) --

Or this file

Expected behaviour
Hoon should check that the second case of $@ is a cell

Notify maintainers
If you happen to know who the appropriate maintainers are, consider mentioning them with an @ here. You may want to use git blame to see who has last touched any relevant code.

[eamsden]

HT @midsum-salrux

[joemfb]

I've looked at this sort of thing a little big. $ runes (specs) are all just macros, expanded prior to type inference, so I think we'll need a natural hoon for setting gol to fix these.

[ohAitch]

I think correct handling of e.g. $@(ship (list)) suggests expanding $@(a b) to $?($>(@ a) $<(@ b)), if I remember what does runes do correctly.

…

On Mon, Nov 14, 2022 at 12:47, Edward Amsden ***@***.***> wrote: *Describe the bug* $@(%our ship) compiles *To Reproduce* Compile |% +$ wat $@(%our ship) *Expected behaviour* Hoon should check that the second case of $@ is a cell *Notify maintainers* If you happen to know who the appropriate maintainers are, consider mentioning them with an @ here. You may want to use git blame to see who has last touched any relevant code. — Reply to this email directly, view it on GitHub <#6081>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOFPBSDY7RM3PJUSXUBWLTWIKQNDANCNFSM6AAAAAASAHDJPQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[joemfb]

Unfortunately $> and $< currently assert relative to the head of a cell, limiting them to specializations of $%. I'm inclined to think they should work as you suggest.

Also, $@ shouldn't need any runtime behavior, just a nest check without a cast or assertion (which is why I'm suggesting a new hoon for gol checking).

[ohAitch]

I mean the way it's documented, it wants runtime behavior to ?- on @ vs ^ when molding, though I do think it expands to just $? in practice.